Variational Smoothing for Pyramids

[pbehne]

Added support to the VariationalMeshSmoother for pyramids.

Ref. #4082

[roystgnr]

I still feel like there's some sort of Clue we're missing, something suboptimal about the variational scheme, that would suggest why our "minimally distorted" higher-order pyramids are still getting distorted in an unexpected way, but for now I'm content to put that up on the "pyramids kinda suck" shelf and add what support we can for them.

[pbehne]

From the failing min clang recipe:

1) test: MeshSmootherTest::testVariationalHex27Tangled (E) 
uncaught exception of type std::exception (or derived).
- Unrecoverable failure to converge

Funny, it converges on my machine :)

[moosebuild]

Job Coverage, step Generate coverage on 290bfcc wanted to post the following:

Coverage

	82cc40	#4232 290bfc
	Total	Total	+/-	New
Rate	64.57%	64.72%	+0.14%	96.97%
Hits	75998	76189	+191	32
Misses	41693	41534	-159	1

Diff coverage report

Full coverage report

This comment will be updated on new commits.

[roystgnr]

The "Test MOOSE" (inc. clang) issues are downstream and unrelated.

The prior "Test with threads" timeouts still worry me. Getting through unit tests in CI opt mode should take like 20 minutes tops; we shouldn't be coming anywhere near a 60 minute timeout. Cutting the mesh sizes from 4^n to 2^n in the case of higher order elements gets us down to 26.5 minutes, which seems fine, but I may kick that a few times before we merge just to see what the variation is.